{"id":"0be6a180-08c1-4b03-ac8c-c6f87f68078e","arxiv_id":"2608.00009","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A benchmark comparing five memory strategies for conversational agents on three public datasets finds dense vector retrieval alone maintains long-range recall, at roughly 25 times the token cost of recency windows.","lead":"AgentMemBench compares five ways AI agents can remember past conversations — recent text, vector search, knowledge graphs, summaries, and web lookup — on three public dialogue datasets. It finds that only vector search reliably recalls facts from many sessions back, but at about 25 times the token cost of the simplest memory strategies.","discovery_kind":"new_method","skeptic_critique":null,"referee_report":null,"author_rebuttal":null,"desk_editor":null,"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-04T01:47:38.494686+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}